intern-lang 1.0.0

Fast string and symbol interning - identifier comparisons become integer comparisons.
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133

//! Criterion benchmarks for the interner hot paths.
//!
//! The two paths that matter most are interning a string that has already been
//! seen (a hash lookup with no allocation — the common case once a file is warm)
//! and interning a fresh string (which appends to the store and may grow the
//! index). Resolution is the read path a symbol table leans on.

use std::hint::black_box;
use std::sync::Arc;
use std::thread;
use std::time::Instant;

use criterion::{BatchSize, Criterion, criterion_group, criterion_main};
use intern_lang::{ConcurrentInterner, Interner};

/// A spread of identifier-shaped strings of varying length.
fn corpus(n: usize) -> Vec<String> {
    (0..n).map(|i| format!("identifier_number_{i}")).collect()
}

fn bench_intern_existing(c: &mut Criterion) {
    let words = corpus(1_024);
    let mut interner = Interner::new();
    for w in &words {
        let _ = interner.intern(w);
    }

    c.bench_function("intern_existing/1024", |b| {
        b.iter(|| {
            for w in &words {
                let _ = black_box(interner.intern(w));
            }
        });
    });
}

fn bench_intern_new(c: &mut Criterion) {
    let words = corpus(1_024);

    c.bench_function("intern_new/1024", |b| {
        b.iter_batched(
            Interner::new,
            |mut interner| {
                for w in &words {
                    let _ = black_box(interner.intern(w));
                }
                interner
            },
            BatchSize::SmallInput,
        );
    });
}

fn bench_resolve(c: &mut Criterion) {
    let words = corpus(1_024);
    let mut interner = Interner::new();
    let symbols: Vec<_> = words.iter().map(|w| interner.intern(w)).collect();

    c.bench_function("resolve/1024", |b| {
        b.iter(|| {
            for &sym in &symbols {
                let _ = black_box(interner.resolve(sym));
            }
        });
    });
}

fn bench_get_miss(c: &mut Criterion) {
    let mut interner = Interner::new();
    for w in &corpus(1_024) {
        let _ = interner.intern(w);
    }
    let absent = corpus(1_024)
        .iter()
        .map(|w| format!("{w}_absent"))
        .collect::<Vec<_>>();

    c.bench_function("get_miss/1024", |b| {
        b.iter(|| {
            for w in &absent {
                let _ = black_box(interner.get(w));
            }
        });
    });
}

/// Throughput of the warm read path (interning strings already present) as the
/// thread count rises. Since hits are served under a shared read lock, this shows
/// how the concurrent interner scales when most identifiers are already known.
fn bench_concurrent_intern_existing(c: &mut Criterion) {
    let words = Arc::new(corpus(1_024));
    let interner = Arc::new(ConcurrentInterner::new());
    for w in words.iter() {
        let _ = interner.intern(w);
    }

    let mut group = c.benchmark_group("concurrent_intern_existing");
    for &threads in &[1usize, 4, 8] {
        group.bench_function(format!("{threads}t/1024"), |b| {
            b.iter_custom(|iters| {
                let start = Instant::now();
                for _ in 0..iters {
                    let handles: Vec<_> = (0..threads)
                        .map(|_| {
                            let interner = Arc::clone(&interner);
                            let words = Arc::clone(&words);
                            thread::spawn(move || {
                                for w in words.iter() {
                                    let _ = black_box(interner.intern(w));
                                }
                            })
                        })
                        .collect();
                    for handle in handles {
                        let _ = handle.join();
                    }
                }
                start.elapsed()
            });
        });
    }
    group.finish();
}

criterion_group!(
    benches,
    bench_intern_existing,
    bench_intern_new,
    bench_resolve,
    bench_get_miss,
    bench_concurrent_intern_existing
);
criterion_main!(benches);